Genetic identification of factors for extracellular cellulose accumulation in the thermophilic cyanobacterium Thermosynechococcus vulcanus: proposal of a novel tripartite secretion system

Tlr1210 has high diguanylate cyclase activity

To distinguish the cellulose synthase system from its regulation by c-di-GMP, we constructed a strain that constitutively overexpressed diguanylate cyclase. In T. vulcanus, nine genes encode GGDEF domains and thus represent potential diguanylate cyclases (Supporting Information Fig. S1). Of these, we chose tlr1210 that encodes a putative active GGDEF domain but no known sensory or signalling domains (Supporting Information Fig. S2 and Fig. 1A). First, we expressed tlr1210 in E. coli and isolated a protein of 42 kDa, which is consistent with the full-length sequence (Fig. 1B). We confirmed its diguanylate cyclase activity by measuring the side product, pyrophosphate (PPi) (Maeda et al., 2014). The specific activity of Tlr1210 was approximately 0.025 μmol c-di-GMP mg protein⁻¹ min⁻¹ (Fig. 1C and D) or approximately 2.26 μmol PPi μmol protein⁻¹ min⁻¹, which is comparable with that reported for the activated form of SesA (1.99 μmol PPi min⁻¹ μmol protein⁻¹ (Enomoto et al., 2014). Moreover, we did not see any product inhibition for Tlr1210, in contrast with the strong product inhibition for SesA, although many residues critical for catalytic and regulatory activities are conserved (Supporting Information Fig. S2). Then, we disrupted tlr1210 by insertional mutation and confirmed the complete segregation of the mutant chromosome by PCR analysis (Supporting Information Fig. S3). The mutant tlr1210 strain had no apparent phenotype such as cell aggregation when compared with the wild type under normal or cell aggregation-inducing conditions (blue light and 31°C) (Supporting Information Fig. S4). These findings suggested that tlr1210 is cryptic but encodes the active enzyme and validated that tlr1210 overexpression may be used to distinguish the cellulose biosynthetic process from its regulation by c-di-GMP.

Effects of overexpression of Tlr1210 and Tll0007

Next, we introduced a construct of tlr1210 with the strong psaA promoter into a neutral region of the genome (Supporting Information Fig. S3). We established the resulting tlr1210 overexpression strain, OX-tlr1210, after complete segregation. In contrast to the wild-type strain, OX-tlr1210 exhibited a stable phenotype of strong cell aggregation at 31°C, which was essentially insensitive to blue-light irradiation (Fig. 2). In the uninduced condition, however, OX-tlr1210 showed only slightly enhanced cell aggregation compared with wild type. These findings suggested that a yet unknown factor(s) is critical for cellulose synthase function under the induced conditions in addition to c-di-GMP. Finally, we introduced the cellulose synthase tll0007 (with a strong promoter) into the OX-tlr1210 strain (Supporting Information Fig. S3) to generate strain OX-tll0007/tlr1210. After complete segregation, OX-tll0007/tlr1210 exhibited lower cell aggregation than did OX-tlr1210.

We also examined the accumulation of extracellular cellulose using fluorescence staining of β-glucan with calcofluor. Under normal growth conditions, the non-aggregated wild-type cells showed very little staining (Fig. 2B and C) and the aggregated cells of OX-tlr1210 showed only weak staining (Fig. 2D and E). Our previous study revealed that the key substance in T. vulcanus cell aggregation is cellulose, which is produced by Tll0007. The extraction method we used therein was complicated and time-consuming (Kawano et al., 2011). In this study, we developed a simpler protocol using ionic liquid extraction of cellulose, which allowed better sample accessibility and more efficient glucose liberation. Indeed, we obtained cellulose yields similar to those generated by our previous laborious method.

When wild-type and OX-tlr1210 cells were grown under normal conditions, OX-tlr1210 cells had a slightly higher cellulose content; however, this amount was significantly less than that generated by the wild-type aggregated cells after induction (31°C with blue light; Fig. 3). The additional overexpression of tll0007 in the OX-tlr1210 mutant did not enhance the cellulose level. These findings suggested that, in addition to cellulose synthase Tll0007 and c-di-GMP, induction generates another unidentified factor(s) that boosts cellulose production.

Synteny analysis of tll0007 homologs in cyanobacterial genomes

In general, the bacterial cellulose synthase gene bcsA is accompanied by related bcs genes. To identify tll0007-linked genes, we analysed synteny among cellulose synthase genes in other cyanobacterial genomes, although only certain species possess these genes. In this case, it is important to recognize orthologs and paralogs, because tlr1795 (cellulose synthase-like gene) is a paralog of tll0007 in the T. vulcanus genome (Kawano et al., 2011). Orthologs of both tll0007 and/or tlr1795 were detected for Nostoc punctiforme ATCC 29133 and Synechococcus elongatus PCC 7942 (S. 7942) (see Supporting Information Table S1). Except for tll0007, they are accompanied by an hlyD-like gene in their neighbourhood (Fig. 4A). Cluster analysis (Fig. 4B) revealed that tll0007 and its orthologs (Npun_F6500 and Synpcc7942_1398) are clustered together, as are the hlyD-like genes (Npun_F6499 and Synpcc7942_1399) for these tll0007 orthologs. Conversely, tlr1795 and its orthologs are separately clustered as distinct groups, as are the corresponding hlyD-like genes. This analysis also revealed that an hlyD-like gene tlr0903 is closely linked to tll0007. Moreover, only the tll0007 orthologs, but not those of tlr1795, are accompanied by endoglucanase-like genes (Npun_F6501 and Synpcc7942_1400). Finally, we found that tlr1902 is closely linked to tll0007. A summary of the cluster analysis results is given in Supporting Information Table S1.

The HlyD domain (Pfam PF16576) is found in HlyD and related proteins called periplasmic adaptors in the tripartite efflux systems of bacteria that include the type I secretion system (Nicaud et al., 1985; Hinchliffe et al., 2013; Kanonenberg et al., 2013). Cellulose fibrils produced by the cellulose synthase on the plasma membrane may be guided through the periplasmic space via an HlyD-like protein in cyanobacteria. Endoglucanase is often an essential factor for cellulose biosynthesis in bacteria and plants, although its precise role is unknown (Nakai et al., 2013). The synteny analysis of the tll0007 orthologs suggested that tlr0903 and tlr1902 may be involved in the biosynthesis and secretion of cellulose that lead to aggregation of T. vulcanus cells.

Physiological roles of tlr0903 and tlr1902

The physiological roles of tlr0903 and tlr1902 were studied by gene inactivation. The deletion mutant genes, Δtlr0903 and Δtlr1902, were segregated from wild-type copies. The cell aggregation phenotypes of the derivative mutant strains were evaluated under the induced conditions of 31°C and blue light (Fig. 5). As expected, the deletion mutations markedly suppressed cell aggregation. These results suggested that tlr0903 and tlr1902 play key roles in cellulose biosynthesis in T. vulcanus.

We then examined the low-temperature induction of these genes by quantitative PCR (qPCR). Total RNA was extracted from T. vulcanus cells at two time points after induction, namely 2 h and 48 h, and transcripts of tll0007, tlr0903 and tlr1902 at 31°C and 45°C were compared (Fig. 6). At both time points, the transcript levels were upregulated at 31°C compared with 45°C. In particular, tlr0903 was markedly upregulated: fivefold at 2 h, and eightfold at 48 h. These results suggested that the transcriptional upregulation of tlr0903 may be another critical event in the induction of cell aggregation and cellulose accumulation in T. vulcanus.

Analysis of tlr0903 and tlr1902 overexpression

We generated various overexpression strains for tlr0903 and tlr1902, the hlyD-like and endoglucanase-like genes respectively. The double overexpression strain, OX-tlr1210/tlr0903, exhibited strong cell aggregation under normal conditions with an aggregation index of 93.3 ± 3.1%, which was much higher than that of the parent strain OX-tlr1210 (57.5 ± 4.1%; Fig. 7A and B). This aggregation level was greater than that of the wild type under induced conditions, that is, 31°C and blue light (aggregation index = 84.3 ± 5.7%). Further deletion of cellulose synthase components (OX-tlr1210/tlr0903 Δtll0007, Supporting Information Fig. S9) suppressed this strong aggregation. The calcofluor staining of β-glucan showed that the aggregated cells of OX-tlr1210/tlr0903 were surrounded with bright blue fluorescence, whereas the non-aggregated cells of the wild type and the additional deletion mutant (OX-tlr1210/tlr0903 Δtll0007) were hardly fluorescent except for a weak red fluorescence that emanated from photosynthetic pigments (Fig. 7C). Treatment with cellulase efficiently dispersed the aggregated cells of OX-tlr1210/tlr0903 (Supporting Information Fig. S7). After extraction with an ionic liquid, we determined that the amount of accumulated cellulose in the aggregated cells of OX-tlr1210/tlr0903 was 7.40 ± 0.52 μg for 4 × 10⁹ cells, which was significantly greater than that determined for the single OX-tlr1210 strain under normal growth conditions of white light at 45°C (Fig. 3). Specifically, the accumulation of cellulose in OX-tlr1210/tlr0903 without induction was nearly equal to, or slightly greater than, that of wild-type cells after induction. Similarly, a single overexpression of tlr0903 (OX-tlr0903) showed stronger cell aggregation than the wild type under blue light at 45°C (Supporting Information Fig. S8). Taking these results and the gene expression analysis into consideration, we concluded that enhanced expression of tlr0903 is a critical event in low temperature-induced cell aggregation. Conversely, OX-tlr0903 showed decreased cell aggregation even compared with that of wild type under white light at 45°C (Fig. 7B). This finding suggested that appropriate balance among the factors might be important for partial cell aggregation.

We also introduced an overexpression construct of tlr1902, which encodes an endoglucanase, in wild type and overexpression strains of tlr1210 and/or tlr0903 to potentially deliver different expression levels of the cellulose synthase system (OX-tlr1902, OX-tlr1210/tlr1902 and OX-tlr1210/tlr0903/tlr1902, respectively). In any case, the cell aggregation or cellulose accumulation was not enhanced, but even suppressed compared with their parent strains (Fig. 7). Conversely, deletion of tlr1902 in the OX-tlr1210/tlr0903 background (OX-tlr1210/tlr0903 Δtlr1902) caused almost no cell aggregation (Fig. 7), a phenotype similar to that exhibited by the single Δtlr1902 mutant after induction (compare to Fig. 5). The cellulose accumulation was also very low in the OX-tlr1210/tlr0903 Δtlr1902 strain (Fig. 3). These results suggested that tlr1902 is indeed one of the key factors for cellulose biosynthesis but that the notion ‘the more, the better’ does not apply to tlr1902. We conclude that the appropriate balance of both the levels and activities of enzymatic components is needed for cellulose biosynthesis.

TolC is coupled with HlyD-like protein

Periplasmic HlyD-like protein is typically linked with an outer membrane efflux protein, TolC, in Gram-negative bacteria (Hinchliffe et al., 2013). A TolC trimer forms a β-barrel pore in the outer membrane and an α-helical domain inward, which is associated with periplasmic adaptor proteins consisting of HlyD domain. There are two tolC homologs (tlr1605 and tll0635) in the Thermosynechococcus genome. Gene inactivation revealed that tlr1605 was only partially inactivated but the cellulose-mediated cell aggregation was totally lost, whereas tll0635 was dispensable for survival and the cell aggregation (Fig. 5 and Supporting Information Fig. S10). This finding supported the idea that TolC family protein Tlr1605 may directly interact with HlyD-like protein Tlr0903 to enable secretion of the nascent cellulose chain across the outer membrane.

Dual regulatory model for cell aggregation in T. vulcanus

We propose a dual regulatory model for the low-temperature/blue–green light-induced aggregation of T. vulcanus cells (Fig. 9). The light signal is recognized by three cyanobacteriochrome-type photoreceptors, namely SesA, SesB and SesC, and transduced into c-di-GMP signalling (Enomoto et al., 2015). SesA is the main photoreceptor that recognizes blue light and produces c-di-GMP. SesB is the antagonistic photoreceptor that recognizes teal light and subsequently degrades c-di-GMP. SesC is a unique dual sensor that, upon stimulation with blue or green light, induces diguanylate cyclase or phosphodiesterase activity respectively. Under blue light, the combination of these three photosensory systems elevates the c-di-GMP level, which in turn activates the cellulose synthase XcsA/Tll0007 via its C-terminal PilZ domain. Nevertheless, even if cells are illuminated with blue light, cell aggregation and cellulose accumulation remain low at the normal temperature. A reduction in temperature induces transcriptional upregulation of xcsB/tlr0903 via yet unknown signalling. It is likely that XcsB/Tlr0903 binds to XcsA/Tll0007 to activate its synthase activity. Thus, cellulose production and cell aggregation are induced only when two independent signals are transmitted simultaneously. Species of Thermosynechococcus and the related Synechococcus are often found as the most abundant phototrophs in microbial mats in hot springs, where the phototrophic niche is very limited (Ward et al., 1998). We assume that the signals-dependent cellulose-mediated cell aggregation is the principal mechanism for the formation of phototrophic microbial mats. A study as to how these mats are formed and maintained under natural conditions is of particular interest from the viewpoint of the regulation of cellulose biosynthesis.

The other novel key factor is the putative endoglucanase XcsC/Tlr1902, classified as Glucoside Hydrolase family 10 (GH_10) and formerly described as the cellulase family F [Carbohydrate-Active enZYmes Database, CAZy, http://www.cazy.org/(Henrissat, 1991; Lombard et al., 2014)]. It is well known that optimal biosynthesis of cellulose requires cellulase activity (endo-1,4-β-d-glucanase). For example, BcsZ (GH_8, formerly cellulase family D) is needed for cellulose biosynthesis in the typical BcsAB-type cellulose synthase. The GH_9 endoglucanase KORRIGAN plays important roles in the trafficking of the cellulose synthase complex in plants (Lei et al., 2014; Vain et al., 2014). Such endoglucanases may contribute to the reorganization and crystallization of nascent cellulose microfibrils (Nakai et al., 2013). Accordingly, we speculate that the putative GH_10 endoglucanase XcsC/Tlr1902 may play a similar and crucial role in the cellulose synthase XcsA/Tll0007.

Phylogenetic analysis and classification of prokaryotic cellulose synthases

We carried out a phylogenetic analysis of the prokaryotic cellulose synthases and related β-glucan synthases using the plant cellulose synthase CesA as the outgroup (Fig. 10). They were separated into at least four major clusters: cluster 1 corresponds to the ‘bacterial’ cellulose synthase consisting of BcsA-BcsB in many proteobacteria and some cyanobacteria (Römling and Galperin, 2015); cluster 2 includes XcsA/Tll0007 (Kawano et al., 2011) and related proteins, which are mainly found in cyanobacteria; cluster 3 corresponds to curdlan ((1→3)-β-d-glucan) synthase and other β-glucan synthases of several proteobacteria (Stasinopoulos et al., 1999; Perez-Mendoza et al., 2015); and cluster 4 includes the cell wall-type cellulose synthase of certain cyanobacteria (Zhao et al., 2015) and related synthases of certain proteobacteria. The PilZ domain for binding of c-di-GMP is found in clusters 1 and 2 but is absent from clusters 3 and 4. Genes of clusters 2 and 3 are accompanied by an hlyD-like gene. HlyD-like bgsB (for cluster 3) was essential for production of the mixed linkage (1→3)(1→4)-β-d-glucan in Sinorhizobium meliloti (Perez-Mendoza et al., 2015). It may suggest that the mixed linkage glucan is secreted through a TolC pore on the outer membrane. According to the phylogeny, the cell wall-type cellulose synthase (cluster 4) first diverged and then the rest further diverged into clusters 2/3 and cluster 1, depending on the pathway to cross the outer membrane (TolC in clusters 2 and 3 vs. BcsC porin in cluster 1). In other words, the phylogenetic divergence of the cellulose synthases of cluster 1 from cluster 2 may reflect the different structural partners of the synthase on its periplasmic side. We must keep in mind that the linkage specificity of these glucan synthases might independently have changed in evolution as seen in genes of cluster 3. Finally, because most cyanobacteria contain genes of clusters 2 and/or 4, the phylogenetically common cellulose synthase genes may have originated in a subset of cyanobacteria and were subsequently transferred to other bacteria.

Plasmid construction

Primers are listed in Supporting Information Table S2. For plasmid construction, see (Maeda et al., 2014; Enomoto et al., 2015). In brief, for overexpression of the diguanylate cyclase Tlr1210 in E. coli: tlr1210 was amplified by PCR from the T. elongatus genomic DNA with primers tlr1210-1F/2R, cloned into an intermediate vector and finally, using NdeI and BamHI, subcloned into pET28a (Novagen, Madison, WI), which encoded N-terminal His-tagged Tlr1210. For overexpression of tlr1210 in T. vulcanus, the tlr1210 fragment was subcloned into a plasmid that harbours a homologous recombination platform near tlr1977 derived from T. vulcanus, a chloramphenicol resistance cassette (Cm^R) and a strong promoter of photosystem I component psaA (P_psaA) in vector pT7Blue (Novagen). We used the corresponding listed primers and generated pT77CAtlr1210. The neutral site in this platform was an HpaI site located in the intergenic region between tlr1977 and tll1978. For overexpression of tll0007 plus tlr1210, tll0007 from T. vulcanus was combined with a kanamycin-resistance cassette, the strong promoter trc for tll0007 and a ribosome-binding site for tlr1210. This construct replaced the psaA promoter in pT77CAtlr1210, resulting in pT77KTtll0007-tlr1210. Inactivation of tlr1210 was carried out as in Supporting Information Fig. S3. Deletion of tll0007 was as described (Enomoto et al., 2015). The overexpression and deletion of tlr0903 and tlr1902 were carried out as shown in Supporting Information Figs S5 and S6, and Table S1. Briefly, the native promoter regions of tlr0903 or tlr1902 were replaced with the synthetic spectinomycin-resistance cassette (Sp^R, Supporting Information Figs S5 and S12) and the trc promoter for overexpression and replaced with a cassette that carried a stop codon (TAG) instead of the initiation codon (ATG) of the target gene for inactivation. All plasmid constructs were confirmed by DNA sequencing.

Protein preparation and Western blotting

For expression and purification of His-tagged Tlr1210, we used E. coli strain C41(DE3) (Enomoto et al., 2014; Maeda et al., 2014). The target protein was induced with 0.1 mM isopropyl-β-d-thiogalactopyranoside and shaking at 20°C overnight. After disruption with a French press, the protein was purified by Ni-affinity column chromatography. Proteins subjected to SDS-PAGE were stained with Coomassie Brilliant Blue R-250 or analysed by western blotting using His-probe (Pierce, Rockford, IL).

Measurement of diguanylate cyclase activity

The diguanylate cyclase activity of Tlr1210 was determined at 25°C by measuring the byproduct, pyrophosphate, using a pyrophosphate assay kit (Molecular Probes, Life Technologies, Carlsbad, CA) as described (Maeda et al., 2014). Each reaction contained 20 mM Tris-HCl (pH7.5), 10 mM MgCl₂ and 100 μM GTP. Protein concentrations were measured by using the Bradford method (Bradford, 1976) with bovine serum albumin as the standard.

Strains and cultures

The thermophilic cyanobacterium T. vulcanus strain RKN (equivalent to NIES-2134) that shows positive phototaxis was cultured at 45°C (normal temperature) or 31°C (low temperature) in BG11 medium (Stanier et al., 1971) with bubbling air containing 1.0% (v/v) CO₂. Cultures were continuously illuminated with white light (white fluorescent lamps, 30 μmol photons m⁻² s⁻¹), red light (red LED, λ_max 634 nm, 30 μmol photons m⁻² s⁻¹) or blue light (blue LED, λ_max 448 nm, 5 μmol photons m⁻² s⁻¹).

Transformation of T. vulcanus

T. vulcanus cells were transformed with plasmid DNAs (Iwai et al., 2004). Antibiotics used for screening and maintaining mutants were 5 μg ml⁻¹ chloramphenicol, 80 μg ml⁻¹ kanamycin, 5 μg ml⁻¹ streptomycin and 10 μg ml⁻¹ spectinomycin. Complete segregation of the transformed DNA in the multicopy genome was confirmed by PCR using the primers listed in Supporting Information Table S2.

Cell aggregation and cellulose accumulation

The cellulose-dependent cell aggregation of T. vulcanus was evaluated as the aggregation index (Kawano et al., 2011). In short, the index representing the ratio of aggregated cells was obtained by the formula: Aggregation index (%) = (OD_total – OD_NA)/OD_total × 100, with OD_NA (NonAggregated) = cell density (OD₇₃₀) after spontaneous precipitation of aggregated cells for 30 min and OD_total = OD₇₃₀ after complete dispersion of the aggregates by treatment with cellulase (Worthington). The cellulose surrounding the cells was stained with the β-glucan-specific fluorescent dye, calcofluor white (Fluorescent Brightener 28, SIGMA); upon excitation with UV light, fluorescence at 440 nm was detected.

Cellulose quantification

The cellulose quantification was as described (Kawano et al., 2011), except that the extraction protocol was improved by selective solubilization with the ionic liquid, N-ethyl-N'-methyl-imidazolium dimethyl phosphonate (Sigma-Aldrich) (Fukaya et al., 2008). Briefly, cells were harvested by centrifugation, washed with Milli-Q water, desiccated and then extracted with the ionic liquid at 45°C for 1.5 h. The cell debris was removed by centrifugation at 20,000 × g for 1 h, and the polysaccharides in the supernatant fraction were precipitated by addition of ethanol [final 83% (v/v)], collected by centrifugation at 20,000 × g for 15 min, and washed with 70% ethanol. The remaining polysaccharide fraction was digested three times overnight with final 4 U ml⁻¹ glucoamylase at 37°C to remove glycogen and then with cellulase (final concentration, 200 U ml⁻¹) at 37°C for 96 h. The released glucose was quantified enzymatically using a glucose assay kit (Glucose CII-test, Wako).

Sequence alignment and phylogram

Protein sequences were obtained from NCBI (http://www.ncbi.nlm.nih.gov/), and CyanoBase [(Fujisawa et al., 2017); http://genome.microbedb.jp/cyanobase]. The domain architectures were searched using the Simple Modular Architecture Research Tool, SMART (Letunic et al., 2015). The classifications of cellulose synthase and glycoside hydrolases were based on CAZy database (http://www.cazy.org/) (Henrissat, 1991; Lombard et al., 2014). Amino acid sequence similarity was evaluated by NCBI BLAST search. Sequence alignments were performed using ClustalX2 (Larkin et al., 2007), and from these, we constructed phylogenetic trees by the bootstrap neighbour-joining method in ClustalX2.

Real-time qPCR

Cells were harvested by centrifugation at 5,000 × g for 10 min at 4°C. We used the RNeasy Mini kit for bacteria (Qiagen) for cell disruption and RNA extraction and mechanical homogenization with zirconia beads (0.1 mm diameter) in a micro-homogenization system (Micro Smash MS-100, Tomy) at 5,000 r.p.m. for 40 s, five times. For cDNA preparation, RNA was reverse transcribed using random primers (PrimeScript RT reagent kit with gDNA eraser, Takara). Real-time qPCR was performed using THUNDERBIRD SYBR qPCR Mix (Toyobo) and the Thermal Cycler Dice Real Time System II (Takara). The expression level of each gene was normalized to the internal control (rnpB) and induction at 31°C was given as a relative value to that at 45°C. The primers are listed in Supporting Information Table S2.